A Graphical Social Topology Model for RGB-D Multi-Person Tracking

Tracking multiple persons is a challenging task especially when move in groups and occlude one another. Existing research have investigated the problems of group division segmentation; however, lacking overall person-group topology modeling limits ability to handle complex person dynamics. We propose Graphical Social Topology (GST) model RGB-D data domain, estimate object dynamics by jointly structure states using topological representation. With our representation, moving are not only assigned groups, but also dynamically connected with each other, which enables in-group individuals be correctively associated cohesion precisely modeled. Using learned typical pattern online update modules, we infer birth/death merging/splitting dynamic groups. GST model, proposed multi-person tracker can naturally facilitate occlusion problem treating occluded other members as whole, while leveraging state transition. Experiments on different RGB datasets confirm that improves state-of-the-arts.